AgentA program that uses AI to take actions on your behalf (like booking tickets or drafting emails).

AI (Artificial Intelligence)Software that can “learn” patterns and make decisions or generate content, instead of just following rigid instructions.

AI EthicsThe study of moral issues in AI, like fairness, privacy, transparency, and responsibility.The “should we” questions, not just the “can we.”

AlgorithmA proven recipe a computer follows to solve a problem.

API (Application Programming Interface)A digital doorway that lets different software systems talk to each other.If websites and apps were restaurants, APIs would be the waiters taking your order to the kitchen.

BiasHidden preferences or skewed patterns in AI results, caused by what it was trained on.

ChatbotAn AI system that interacts with people in conversation form.The “face” of an AI that talks back.

EmbeddingsNumbers that represent words, images, or concepts so AI can compare and find similarities.

Fine-TuningAdjusting an AI model after initial training to specialize it for a certain task or industry.Like taking a trained musician and coaching them only in jazz.

Generative AIAI that doesn’t just analyze but creates new content — text, images, audio, or video.A system that makes, not just measures.

HallucinationWhen AI confidently makes something up that isn’t true.

JSONA simple file format for storing and sharing structured data.

Large Language Model (LLM)A type of AI trained on huge amounts of text to predict and generate words in sequence.Like an autocomplete system, but trained on billions of conversations.

LatencyThe delay between asking AI a question and getting its response.Lag time — like waiting for a text to come through.

Machine Learning (ML)A branch of AI where computers learn from examples and improve with experience.

ModelThe trained brain of an AI. It’s what the system has “learned” from massive amounts of data.

Neural NetworkA design for AI inspired by the human brain, with layers of digital “neurons” that pass signals to each other.

Open SourceSoftware whose code is made public so anyone can use, study, or improve it.A community cookbook, instead of a secret family recipe.

PromptThe input you give an AI to get a response (a question, instruction, or piece of text).

Proprietary AISoftware owned and controlled by a company, usually kept closed-source.The locked vault version of AI.

RAG (Retrieval-Augmented Generation) A method where AI fetches fresh facts from a database or the web before answering, instead of only relying on what it was trained on.

Schema A structured way to organize data so AI (and search engines) can understand it better.

Supervised Learning A kind of machine learning where the training data includes correct answers, so the AI learns by example. Like a teacher marking homework.

Synthetic Data Data artificially created (by humans or machines) to train AI models. Like practice problems invented for a student before the real test.

Token The smallest chunk of text an AI reads or writes. It might be a word, part of a word, or even punctuation. Think of tokens as Lego bricks that snap together into sentences.

Training Data The examples fed into an AI to teach it how to respond.

Unsupervised Learning Machine learning without labeled answers — the AI just looks for patterns in raw data. Like sorting a box of puzzle pieces without knowing what the picture looks like.

User Agent Software that acts on behalf of a user, like a browser or chatbot interface that helps manage tasks and interactions.